The Princeton GFDL coupled ocean–atmosphere global climate model (GCM) of greenhouse warming yields estimates of ocean temperature change from 1990–2000 as a function of position and depth. These estimates have served as a basis for computing changes in acoustic travel time between Heard Island and various receiver sites. (I am greatly indebted to S. Manabe for computing these changes.) The results differ from ocean basin to ocean basin and from gyre to gyre. A typical result is a reduction in travel time by 2 s per decade over a 10 000-km path. Mesoscale eddies produce a month-to-month variability of 0.5 s rms. Gyre and basin variability are more troublesome. The goal is to deploy a reasonable network of acoustic sources and receivers with gyre scale resolution and mesoscale suppression.